fix(metrics): align fleet series by time bucket, harden fleet page inputs

Address review findings on the fleet metrics page.

Data correctness:
- bucket raw Sentinel samples into epoch-aligned windows (5m/30m/2h per range)
  and average them, replacing per-server LTTB. Every server now reports the
  same bucket timestamps, so the fleet sums add up point by point instead of
  zigzagging between servers (and RX+TX really combine)
- keep only the root mount in disk history, matching HasMetrics::getDiskMetrics
- leave sparkline gaps as null instead of 0, so absent samples read as gaps

Robustness:
- normalize containerMetric in mount()/updatedContainerMetric(), so
  /metrics?cmetric=foo no longer throws an undefined array key
- include the server uuid in the container row wire:key; system containers
  share one Sentinel id across servers
- round the range start to the minute so the 60s history cache can hit
- cap the Sentinel curl with --max-time 10, so one hung request cannot stall
  the serial per-server loop

UI:
- pass the spark time axis on first paint, so tooltips no longer show 1970
- show the active series name (RX/TX) in the server network chart tooltip
This commit is contained in:
Aditya Tripathi
2026-09-18 07:44:22 +00:00
parent ab436942c9
commit faca466cd6
9 changed files with 163 additions and 54 deletions
@@ -25,12 +25,12 @@ class ChartStubClient extends SentinelMetricsClient
return [];
}
public function history(string $metric, string $from): array
public function history(string $metric, string $range): array
{
return [[1000, 10.0], [2000, 20.0]];
}
public function networkHistory(string $from): array
public function networkHistory(string $range): array
{
return ['rx' => [[1000, 5.0]], 'tx' => [[1000, 1.0]]];
}
@@ -46,7 +46,7 @@ class ChartTestableMetrics extends Metrics
class DenseChartStubClient extends ChartStubClient
{
public function history(string $metric, string $from): array
public function history(string $metric, string $range): array
{
// 300 points, like a downsampled 24h host series.
return array_map(fn ($i) => [$i * 1000, (float) $i], range(1, 300));
@@ -99,12 +99,12 @@ it('includes KPI-tile spark series aligned to one shared time axis', function ()
$payload = $params[0];
// Union of every series' buckets, sorted; each spark projects onto it,
// filling gaps with 0 (network only has the 1000 bucket → tail is 0).
// leaving gaps as null (network only has the 1000 bucket → tail is null).
return $payload['sparkCategories'] == [1000, 2000]
&& $payload['cpuSpark'] == [10, 20]
&& $payload['memSpark'] == [10, 20]
&& $payload['diskSpark'] == [10, 20]
&& $payload['netSpark'] == [6, 0];
&& $payload['netSpark'] === [6, null];
});
});
@@ -30,12 +30,12 @@ class StubMetricsClient extends SentinelMetricsClient
return self::$containers[$this->server->uuid] ?? [];
}
public function history(string $metric, string $from): array
public function history(string $metric, string $range): array
{
return [];
}
public function networkHistory(string $from): array
public function networkHistory(string $range): array
{
return ['rx' => [], 'tx' => []];
}
@@ -189,3 +189,41 @@ it('renders container metric values without leaking a blade directive', function
->assertDontSee('@break')
->assertSee('/s');
});
it('falls back to cpu for an unknown container metric', function () {
$s = metricsServer($this->team, $this->privateKey);
StubMetricsClient::$summaries = [$s->uuid => onlineSummary()];
StubMetricsClient::$containers = [$s->uuid => [
['id' => 'a', 'cpu' => 5.0, 'memUsed' => 900, 'memPercent' => 9.0, 'diskBytes' => 1, 'net' => 0.0],
]];
Livewire::withQueryParams(['cmetric' => 'foo'])
->test(TestableMetrics::class)
->assertOk()
->assertSet('containerMetric', 'cpu')
->set('containerMetric', 'bar')
->assertSet('containerMetric', 'cpu');
});
it('keys container rows by server so shared container ids do not collide', function () {
$one = metricsServer($this->team, $this->privateKey);
$two = metricsServer($this->team, $this->privateKey);
$proxy = ['id' => 'coolify-proxy', 'cpu' => 1.0, 'memUsed' => 10, 'memPercent' => 1.0, 'diskBytes' => 1, 'net' => 0.0];
StubMetricsClient::$summaries = [$one->uuid => onlineSummary(), $two->uuid => onlineSummary()];
StubMetricsClient::$containers = [$one->uuid => [$proxy], $two->uuid => [$proxy]];
Livewire::test(TestableMetrics::class)
->assertSeeHtml('wire:key="fleet-container-'.$one->uuid.'-coolify-proxy"')
->assertSeeHtml('wire:key="fleet-container-'.$two->uuid.'-coolify-proxy"');
});
it('passes the spark time axis to the KPI sparklines on first paint', function () {
$s = metricsServer($this->team, $this->privateKey);
StubMetricsClient::$summaries = [$s->uuid => onlineSummary()];
$component = Livewire::test(TestableMetrics::class);
$component->set('chartData', ['sparkCategories' => [1700000100000], 'cpuSpark' => [12.5]]);
$component->assertSeeHtml('1700000100000');
});
@@ -2,6 +2,7 @@
use App\Models\Server;
use App\Services\SentinelMetricsClient;
use Illuminate\Support\Carbon;
class FakeMetricsClient extends SentinelMetricsClient
{
@@ -93,26 +94,61 @@ it('returns an empty container list when the endpoint 404s', function () {
expect($client->containersCurrent())->toBe([]);
});
it('maps cpu history to [time, value] pairs', function () {
it('averages cpu history into range-aligned buckets', function () {
$client = fakeMetricsClient();
$client->bodies = [
'/cpu/history' => json_encode([
['time' => 1000, 'percent' => '10'],
['time' => 2000, 'percent' => '20'],
['time' => '1700000000000', 'percent' => '10'],
['time' => '1700000060000', 'percent' => '20'],
['time' => '1700000300000', 'percent' => '40'],
]),
];
expect($client->history('cpu', 'from'))->toBe([[1000, 10.0], [2000, 20.0]]);
// 24h range → 5-minute buckets floored to absolute epoch boundaries.
expect($client->history('cpu', '24h'))->toBe([[1699999800000, 15.0], [1700000100000, 40.0]]);
});
it('downsamples long history series so the chart payload stays small', function () {
it('aligns buckets across servers whose samples have different timestamps', function () {
$a = fakeMetricsClient();
$a->bodies = ['/cpu/history' => json_encode([['time' => 1700000001234, 'percent' => '10']])];
$b = fakeMetricsClient();
$b->bodies = ['/cpu/history' => json_encode([['time' => 1700000047890, 'percent' => '30']])];
expect(array_column($a->history('cpu', '24h'), 0))->toBe(array_column($b->history('cpu', '24h'), 0));
});
it('keeps long history series small', function () {
$rows = [];
for ($i = 0; $i < 5000; $i++) {
$rows[] = ['time' => $i, 'percent' => (string) ($i % 100)];
// 24h of raw 20-second samples.
for ($i = 0; $i < 4320; $i++) {
$rows[] = ['time' => 1700000000000 + $i * 20_000, 'percent' => (string) ($i % 100)];
}
$client = fakeMetricsClient();
$client->bodies = ['/cpu/history' => json_encode($rows)];
expect(count($client->history('cpu', 'from')))->toBe(300);
expect(count($client->history('cpu', '24h')))->toBeLessThanOrEqual(300);
});
it('keeps only the root mount in disk history', function () {
$client = fakeMetricsClient();
$client->bodies = [
'/disk/history' => json_encode([
['time' => 1700000000000, 'mount' => '/', 'usedPercent' => 85],
['time' => 1700000000000, 'mount' => '/boot', 'usedPercent' => 20],
]),
];
expect($client->history('disk', '24h'))->toBe([[1699999800000, 85.0]]);
});
it('rounds the range start to the minute so history responses are cacheable', function () {
Carbon::setTestNow('2026-01-01 12:30:05');
$first = SentinelMetricsClient::rangeFrom('24h');
Carbon::setTestNow('2026-01-01 12:30:45');
expect(SentinelMetricsClient::rangeFrom('24h'))->toBe($first)->toBe('2025-12-31T12:30:00Z');
Carbon::setTestNow();
});